This is an advance summary of a forthcoming article in the Oxford Research Encyclopedia of Climate Science. Please check back later for the full article.The concept of statistical downscaling or empirical-statistical downscaling became a distinct and important scientific approach in climate science in recent decades, when the climate change issue and assessment of climate change impact on various social and natural systems have become international challenges. Global climate models are the best tools for estimating future climate conditions. Even if improvements can be made in state-of-the art global climate models, in terms of spatial resolution and their performance in simulation of climate characteristics, they are still skillful only in reproducing large-scale feature of climate variability, such as global mean temperature or various circulation patterns (e.g., the North Atlantic Oscillation). However, these models are not able to provide reliable information on local climate characteristics (mean temperature, total precipitation), especially on extreme weather and climate events. The main reason for this failure is the influence of local geographical features on the local climate, as well as other factors related to surrounding large-scale conditions, the influence of which cannot be correctly taken into consideration by the current dynamical global models.Impact models, such as hydrological and crop models, need high resolution information on various climate parameters on the scale of a river basin or a farm, scales that are not available from the usual global climate models. Downscaling techniques produce regional climate information on finer scale, from global climate change scenarios, based on the assumption that there is a systematic link between the large-scale and local climate. Two types of downscaling approaches are known: a) dynamical downscaling is based on regional climate models nested in a global climate model; and b) statistical downscaling is based on developing statistical relationships between large-scale atmospheric variables (predictors), available from global climate models, and observed local-scale variables of interest (predictands).Various types of empirical-statistical downscaling approaches can be placed approximately in linear and nonlinear groupings. The empirical-statistical downscaling techniques focus more on details related to the nonlinear models—their validation, strengths, and weaknesses—in comparison to linear models or the mixed models combining the linear and nonlinear approaches. Stochastic models can be applied to daily and sub-daily precipitation in Romania, with a comparison to dynamical downscaling. Conditional stochastic models are generally specific for daily or sub-daily precipitation as predictand.A complex validation of the nonlinear statistical downscaling models, selection of the large-scale predictors, model ability to reproduce historical trends, extreme events, and the uncertainty related to future downscaled changes are important issues. A better estimation of the uncertainty related to downscaled climate change projections can be achieved by using ensembles of more global climate models as drivers, including their ability to simulate the input in downscaling models. Comparison between future statistical downscaled climate signals and those derived from dynamical downscaling driven by the same global model, including a complex validation of the regional climate models, gives a measure of the reliability of downscaled regional climate changes.
CliGAN: A Structurally Sensitive Convolutional Neural Network Model for Statistical Downscaling of Precipitation from Multi-Model Ensembles
